PLASMA compiler emits broken module if a constant is defined before imports

[ZornsLemma]

If I modify test.pla to add a constant definition before the includes:

diff --git a/src/samplesrc/test.pla b/src/samplesrc/test.pla
index 73a29eb..a27c27c 100755
--- a/src/samplesrc/test.pla
+++ b/src/samplesrc/test.pla
@@ -1,6 +1,7 @@
 //
 // Include all imported modules and their data/functions.
 //
+const foo = 42
 include "inc/cmdsys.plh"
 include "inc/testlib.plh"
 //

the PLASMA compiler doesn't complain but it generates a broken module with the "end of module dependencies" zero byte too early (before CMDSYS, in this case):

; ACME COMPATIBLE OUTPUT
        !WORD   _SEGEND-_SEGBEGIN       ; LENGTH OF HEADER + CODE/DATA + BYTECODE SEGMENT
_SEGBEGIN
        !WORD   $6502                   ; MAGIC #
        !WORD   _SYSFLAGS                       ; SYSTEM FLAGS
        !WORD   _SUBSEG                 ; BYTECODE SUB-SEGMENT
        !WORD   _DEFCNT                 ; BYTECODE DEF COUNT
        !WORD   _INIT                   ; MODULE INITIALIZATION ROUTINE
; <stdin>: 0001: //
; <stdin>: 0002: // Include all imported modules and their data/functions.
; <stdin>: 0003: //
; <stdin>: 0004: const foo = 42
        !BYTE   $00                     ; END OF MODULE DEPENDENCIES
                                        ; foo = 42
; <stdin>: 0005: include "inc/cmdsys.plh"
; inc/cmdsys.plh: 0001: import cmdsys
        ; DCI STRING: CMDSYS
        !BYTE   $C3,$CD,$C4,$D3,$D9,$53
                                        ; CMDSYS -> X000

I think this should either work or the compiler should refuse to compile the file because there's a constant definition in the wrong place.

[dschmenk]

I tried this with the latest (BETA) and it seems to work fine now.